gms | German Medical Science

Hearing aids have greatly evolved over the last decades, largely due to improvements in the area of digital signal processing. For commercial hearing aids, the principles and algorithms used in their features and how they interact are company secrets and therefore largely unknown to the users or independent researchers. While working with “black box” devices is no major issue for the users or dispensers of hearing aids, it is a significant problem for researchers who want to draw meaningful conclusions from studies using these devices. Further, laboratory experiments with oversimplified setups often do not reflect the real-life effect of hearing aids [1]. Hearing aid research therefore needs devices that are usable in the field, fully controllable by any researcher, and provide the full processing chain of modern, non-linear and adaptive hearing aids.

To pursue this need, since 2015 the National Institute on Deafness and Other Communication Disorders (USA) has posed two requests for funding to develop portable signal processing tools that provide substantial computing power for real-time processing of the perceived acoustic environment. In response to this call, several groups have been developing platforms for hearing aid research that allow for free user programming and mobile use in the field [2], [3], [4], [5]. One example for such research platforms is the Portable Hearing Lab (PHL) [3]. The PHL consists of an extended portable miniature computer with audio interface, tailored Behind-The-Ear with Receiver-In-Canal (BTE-RIC) or In-The-Ear (ITE) hearing aid headsets, and a real-time operating system that includes the open Master Hearing Aid (openMHA), an open-source software framework for hearing aid signal processing [6], [7]. The PHL thus offers a relatively simple possibility to set up a full hearing aid processing chain and run it in real time in portable hardware with an electro-acoustic front-end very comparable to commercial hearing aids.

In this work, we assess standard technical performance metrics of the PHL with its BTE headset according to IEC 60118-0:2015 (largely equivalent to ANSI S3.22:2014 [8]), and compare it to current state-of-the-art hearing aids. With these tests we evaluate whether the basic electro-acoustic performance of the PHL is on a comparable level as that of commercial hearing aids. We purposely refrain from obtaining performance metrics that depend more severely on the signal processing chain, since with the PHL it is intended that the user can configure it very freely to their needs.

Three samples of the PHL were examined (first version, from Beta phase 2020/2021), each comprising one processor, a wired binaural BTE-RIC hearing aid headset, and a set of exchangeable M (medium power output, type Sonion RIC-E50D) and S (small power output, type Sonion RIC-4400) receivers. Measurements were only conducted for the right side of each unit. The latest firmware as of August 2021 (Mahalia 4.16-R2) that included a generic hearing aid configuration of the openMHA was used without modifications. All its algorithms were deactivated except for a five-band dynamic range compressor, into which an equal linear gain was programmed in each channel. Other than for commercial hearing aids, the Full-On-Gain (FOG) setting, i.e., the maximum gain allowed by the fitting software, is not specified for the PHL, but required for measurements according to standards. We defined the FOG by taking the nominal peak output values (included for calibration purposes) and subtracting 65 dB, with the intention that the device would be able to reproduce a speech-like input signal with a crest factor of 15 dB at 50 dB SPL without triggering peak limiting. This resulted in broadband gains of 51 dB with the M receiver, and 45 dB with the S receiver.

For reference, the typical (10% to 90% interquantile) range of measured values of commercial BTE-RIC hearing aids that have undergone type approval in our lab during the year 2020 are given. This includes most BTE-RIC hearing aids that entered the German hearing aid market in 2020 (125 with S receiver, and 130 with M receiver), making up a large coverage of the world market.

Standard electroacoustic performance metrics according to IEC 60118-0:2015 [9] were measured using our test stand in an anechoic chamber. The test stand comprised a 2 cm³ coupler (Brüel & Kjær 4969, with 4192 microphone and 2669 preamplifier) to which the receiver unit of the PHL earpiece was coupled using airtight seal, and a pressure microphone (Brüel & Kjær 4192 with 2669 preamplifier) mounted 12±2 mm above the center point between the hearing aid microphones. Frequency-dependent measurements of the levels in free field and in the coupler with different settings of the PHL were performed using sine tones with variable frequency between 100 and 10,000 Hz. Sound was presented from a loudspeaker positioned in 1 m distance from the device under test, all controlled by a Rohde & Schwarz UPV Audio Analyzer. These responses were captured with the devices under test in FOG setting at 50 dB SPL input (FOG response curve) and 90 dB SPL input (OSPL90). From these measurements, a device-specific Reference Test Gain (RTG) was derived according to IEC 60118-0:2015 and programmed into the devices for measurement of the Equivalent Input Noise (EIN) and Total Harmonic Distortion (THD).

The top panel of Figure 1 [Fig. 1] shows the OSPL90 for each unit of the PHL, as well as the range of appropriate commercial devices. With the M receiver, the PHL reaches an OSPL90 of up to 115 dB SPL peak, and high-frequency average (HFA, across 1,000, 1,600 and 2,500 Hz) of 111.5 dB SPL on average across devices. With the S receiver, these values are lower, with a peak of up to 112 dB SPL and an HFA value of 106.3 dB SPL. The OSPL90 for the PHL lies well within the range or at the top end of appropriate commercial devices with either receiver. Differences between devices reach up to 3 dB and most probably originate from the receivers, which we tested by further measurements where the drivers were exchanged between PHL devices. It can also be seen in Figure 2 [Fig. 2] that the differences between units are not consistent between receiver types.

The bottom panel of Figure 1 [Fig. 1] shows the estimated insertion gain values for frontal incidence, which were computed by subtracting the appropriate CORFIG values [10] as given by Bentler & Pavlovic [11] from the coupler gains (level in coupler re. free field) with FOG setting. The Coupler Response for Flat Insertion Gain (CORFIG) includes corrections for difference in response in the 2 cm³ coupler versus the real ear, as well as effects of microphone location present in the real ear but not the test setup. The PHL with the M receiver achieves insertion gain values around 50 dB, except for a 10 dB dip in the frequency range between 3 and 7 kHz. These values and frequency dependencies are similar to those of commercial devices. The PHL with the S receiver reaches insertion gain values up to 50 dB at frequencies below 1 kHz, and frequency-dependent values between 37 and 50 dB at higher frequencies. Depending on the frequency, these gains are at the upper end or, especially for frequencies lower than 1 kHz, higher than seen in commercial devices.

The target maximum gain settings defined in the present study (shown as dashed lines in the lower panel of Figure 1 [Fig. 1]) were largely met by the PHL within ±3 dB up to 3 kHz in case of the M receiver, but show consistent deviations of up to 5 dB with the S receiver. The estimated insertion gain values with the PHL do not differ more than 1 dB between devices up to 1 kHz, and 4 dB up to 8 kHz, which are smaller variations than for the OSPL90.

Table 1 [Tab. 1] shows a summary of performance values, namely the HFA of the OSPL90 and the coupler gain in FOG setting. Please note that the FOG coupler gains differ from the estimated insertion gain values, and are provided here for reference purposes. Furthermore, the EIN with the M receiver is close to the average of commercial reference devices, while with the S receiver, it is approx. 3 dB above the average but still within one standard deviation of commercial reference devices. It should be noted here that level expansion at low input levels was disabled in the PHL, but may have been active in the commercial devices. The THD is not routinely assessed with type approval of hearing aids in Germany and thus not available for the reference devices. However, the THD values of the PHL are well within the range acceptable for the appropriate receiver types and in the typical to low range of hearing aid datasheet values.

The insertion gain values provided by the PHL were similar to those of commercial devices and correspond well to the programmed gain values with the M receiver, whereas some deviations to the programmed gains were observed with the S receivers. Although the insertion gain values estimated here are only a coarse estimation for an average ear [10], [12], our results demonstrate that as in regular hearing aid fitting, the programmed gain values of the PHL are not necessarily equal to the actual gains achieved in the real ear. Both systematic as well as random deviations from the programmed gain values are to be expected in real ears. The results seen here are well in line with previous measurements on one PHL unit in our lab using a test box setup as well as in KEMAR, where similar or larger deviations to programmed gains were noted in one commercial reference device [13]. The PHL showed higher gains and output level, lower EIN, lower THDs and better match to programmed gain with the M receiver as compared to the S receiver. The differences in output performance between S and M receivers observed in the present study is consistent with the difference in sensitivity provided by the manufacturer. Therefore, if space in the ear canal allows, for most applications the PHL will probably reach better performance when the M receiver is used.

Finally, we conclude that the PHL provides a suitable basis for setting up a full research hearing aid whose basic electro-acoustic performance is equivalent to commercial devices.

The authors declare that they have no competing interests.

We thank Chas Pavlovic for partly providing the PHL units used in this work.